Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.1 Outline Introduction Background Distributed DBMS Architecture Distributed Database Design Distributed Query Processing Distributed Transaction Management n Building Distributed Database Systems (RAID) Mobile Database Systems Privacy, Trust, and Authentication Peer to Peer Systems

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.2 Useful References Y. Lu, W. Wang, D. Xu, and B. Bhargava, Trust-Based Privacy Preservation for Peer-to- peer, in the 1st NSF/NSA/AFRL workshop on secure knowledge management (SKM), Buffalo, NY, Sep

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.3 Problem statement Privacy in peer-to-peer systems is different from the anonymity problem Preserve privacy of requester A mechanism is needed to remove the association between the identity of the requester and the data needed

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.4 Proposed solution A mechanism is proposed that allows the peers to acquire data through trusted proxies to preserve privacy of requester  The data request is handled through the peer’s proxies  The proxy can become a supplier later and mask the original requester

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.5 Related work Trust in privacy preservation  Authorization based on evidence and trust  Developing pervasive trust Hiding the subject in a crowd  K-anonymity  Broadcast and multicast

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.6 Related work (2) Fixed servers and proxies  Publius Building a multi-hop path to hide the real source and destination  FreeNet  Crowds  Onion routing

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.7 Related work (3)  provides sender-receiver anonymity by transmitting packets to a broadcast group Herbivore  Provides provable anonymity in peer-to-peer communication systems by adopting dining cryptographer networks

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.8 Privacy measurement A tuple is defined to describe a data acquirement. For each element, “0” means that the peer knows nothing, while “1” means that it knows everything. A state in which the requester’s privacy is compromised can be represented as a vector, (y Є [0,1]) from which one can link the ID of the requester to the data that it is interested in.

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.9 For example, line k represents the states that the requester’s privacy is compromised. Privacy measurement (2)

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.10 Mitigating collusion An operation “*” is defined as: This operation describes the revealed information after a collusion of two peers when each peer knows a part of the “secret”. The number of collusions required to compromise the secret can be used to evaluate the achieved privacy

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.11 Trust based privacy preservation scheme The requester asks one proxy to look up the data on its behalf. Once the supplier is located, the proxy will get the data and deliver it to the requester  Advantage: other peers, including the supplier, do not know the real requester  Disadvantage: The privacy solely depends on the trustworthiness and reliability of the proxy

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.12 Trust based scheme – Improvement 1 To avoid specifying the data handle in plain text, the requester calculates the hash code and only reveals a part of it to the proxy. The proxy sends it to possible suppliers. Receiving the partial hash code, the supplier compares it to the hash codes of the data handles that it holds. Depending on the revealed part, multiple matches may be found. The suppliers then construct a bloom filter based on the remaining parts of the matched hash codes and send it back. They also send back their public key certificates.

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.13 Trust based scheme – Improvement 1 Examining the filters, the requester can eliminate some candidate suppliers and finds some who may have the data. It then encrypts the full data handle and a data transfer key k data with the public key. The supplier sends the data back using k data through the proxy Advantages:  It is difficult to infer the data handle through the partial hash code  The proxy alone cannot compromise the privacy  Through adjusting the revealed hash code, the allowable error of the bloom filter can be determined

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.14 Data transfer procedure after improvement 1 R: requester S: supplier Step 1, 2: R sends out the partial hash code of the data handle Step 3, 4: S sends the bloom filter of the handles and the public key certificates Step 5, 6: R sends the data handle and encrypted by the public key Step 7, 8: S sends the required data encrypted by Requester Proxy of Supplier Requester

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.15 Trust based scheme – Improvement 2 The above scheme does not protect the privacy of the supplier To address this problem, the supplier can respond to a request via its own proxy

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.16 Trust based scheme – Improvement 2 Requester Proxy of Proxy of Supplier Requester Supplier

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.17 Trustworthiness of peers The trust value of a proxy is assessed based on its behaviors and other peers’ recommendations Using Kalman filtering, the trust model can be built as a multivariate, time-varying state vector

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.18 Experimental platform - TERA Trust enhanced role mapping (TERM) server assigns roles to users based on  Uncertain & subjective evidences  Dynamic trust Reputation server  Dynamic trust information repository  Evaluate reputation from trust information by using algorithms specified by TERM server

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.19 Trust enhanced role assignment architecture (TERA)

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.20 Conclusion A trust based privacy preservation method for peer- to-peer data sharing is proposed It adopts the proxy scheme during the data acquirement Extensions  Solid analysis and experiments on large scale networks are required  A security analysis of the proposed mechanism is required

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.21 Peer to Peer Systems and Streaming

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.22 Useful References G. Ding and B. Bhargava, Peer-to-peer File-sharing over Mobile Ad hoc Networks, in the First International Workshop on Mobile Peer-to-Peer Computing, Orlando, Florida, March M. Hefeeda, A. Habib, B. Botev, D. Xu, and B. Bhargava, PROMISE: Peer-to-Peer Media Streaming Using CollectCast, In Proc. of ACM Multimedia 2003, 45-54, Berkeley, CA, November 2003.

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.23 Overview of Peer-to-Peer (P2P) Systems Peer  Autonomy: no central server  Similar power Share resources among a large number of peers P2P is a distributed system where peers collaborate to accomplish tasks

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.24 P2P Applications P2P file-sharing  Napster, Gnutella, KaZaA, eDonkey, etc. P2P Communication  Instant messaging  Mobile Ad hoc network P2P Computation 

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.25 P2P Searching Algorithms Search for file, data, or peer Unstructured  Napster, Gnutella, KaZaA, eDonkey, etc. Structured  Chord, Pastry, Tapestry, CAN, etc.

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.26 Napster: Central Directory Server Bob wants to contact Alice, he must go through the central server Benefits:  Efficient search  Limited bandwidth usage  No per-node state Drawbacks:  Central point of failure  Limited scale  Copyrights BobAlice JaneJudy Peer Central Server

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.27 Gnutella: Distributed Flooding Bob wants to talk to Alice, he must broadcast request and get information from Jane Benefits:  No central point of failure  Limited per-node state Drawbacks:  Slow searches  Bandwidth intensive  Scalability Bob Alice Jane Judy Carl

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.28 KaZaA: Hierarchical Searching Bob talks to Alice via Server B and Server A. Popularity:  More than 3 M peers  Over 3,000 Terabytes  >50% Internet traffic ? Benefits:  Only super-nodes do searching  Parallel downloading  Recovery Drawbacks:  Copyrights Bob Alice SB SA

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.29 Peers characterized as  Highly diverse  Dynamic  Have limited capacity, reliability Problem  How to select and coordinate multiple peers to render the best possible quality streaming? P2P Streaming

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.30 CollectCast (Developed at Purdue) CollectCast is a new P2P service  Middleware layer between a P2P lookup substrate and applications  Collects data from multiple senders Functions  Infer and label topology  Select best sending peers for each session  Aggregate and coordinate contributions from peers  Adapt to peer failures and network conditions

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.31 CollectCast (cont’d)

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.32 Simulations Compare selection techniques in terms of  The aggregated received rate, and  The aggregated loss rate  With and without peer failures Impact of peer availability on size of candidate set Size of active set Load on peers

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.33 Simulation: Setup Topology  On average 600 routers and 1,000 peers  Hierarchical (Internet-like) Streaming session  Rate R 0 = 1 Mb/s  Duration = 60 minutes  Loss tolerance level α u = 1.2 Peers  Offered rate: uniform in [0.125R 0, 0.5R 0 ]  Availability: uniform in [0.1, 0.9]  Diverse P2P community Results are averaged over 100 runs with different seeds

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.34 Aggregate Rated: No Failures Careful selection pays off!

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.35 PROMISE and Experiments on PlanetLab (Test-bed at Purdue) PROMISE is a P2P media streaming system built on top of CollectCast Tested in local and wide area environments Extended Pastry to support multiple peer look up

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.36 PlanetLab Experiments PROMISE is installed on 15 nodes Use several MPGE-4 movie traces Select peers using topology-aware (the one used in CollectCast) and end-to-end Evaluate  Packet-level performance  Frame-level performance and initial buffering  Impact of changing system parameters  Peer failure and dynamic switching

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.37 Packet-Level: Aggregated Rate Smoother aggregated rate achieved by CollectCast

Distributed DBMS© 2001 M. Tamer Özsu & Patrick Valduriez Page 0.38 Conclusions New service for P2P networks (CollectCast)  Infer and leverage network performance information in selecting and coordinating peers PROMISE is built on top of CollectCast to demonstrate its merits Internet Experiments show proof of concept  Streaming from multiple, heterogeneous, failure-prone, peers is indeed feasible Extend P2P systems beyond file sharing Concrete example of network tomography